SEARCH · PubMed Health

Results for “Predictive Learning Models”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5Linked to original sources

PubMed2026

Prediction of bacterial protein-compound interactions with only positive samples.

MOTIVATION: Prediction of Compound-Protein Interactions (CPI) in bacteria is crucial to advance various pharmaceutical and chemical engineering fields, including biocatalysis, drug discovery, and industrial processing. However, current CPI models cannot be applied for bacterial CPI prediction due to the lack of curated negative interaction samples. RESULTS: We propose a novel Positive-Unlabeled (PU) learning framework, named BIN-PU, to address this limitation. BIN-PU generates pseudo positive and negative labels from known positive interaction data, enabling effective training of deep learning models for CPI prediction. We also propose a weighted positive loss function that weights to truly positive samples. We have validated BIN-PU coupled with multiple CPI backbone models, comparing the performance with the existing PU models using bacterial cytochrome P450 (CYP) data. Extensive experiments demonstrate the superiority of BIN-PU over the benchmark models in predicting CPIs with only truly positive samples. Furthermore, we have validated BIN-PU on additional bacterial proteins obtained from literature review, human CYP datasets, and uncurated data for its reproducibility. We have also validated the CPI prediction for the uncurated CYP data with biological and biophysical experiments. BIN-PU represents a significant advancement in CPI prediction for bacterial proteins, opening new possibilities for improving predictive models in related biological interaction tasks. AVAILABILITY AND IMPLEMENTATION: The source code and data are available at https://github.com/datax-lab/CYP.

Bacterial Proteins↗

PubMed2026

Investigating cross-organism prediction of prokaryotic essential proteins using unsupervised language model and ensemble strategy.

Cross-organism prediction of essential proteins is a critical task for drug discovery and microbial engineering, yet the generalizability of existing machine learning models across diverse species remains a significant challenge. In this study, we propose DeepPEP, a large language model-based framework designed to reliably transfer essential protein annotations between distantly related organisms. Utilizing 66 curated prokaryotic datasets, we systematically evaluated DeepPEP's cross-organism performance under various conditions. Initial pairwise predictions revealed a correlation between performance and evolutionary distance; however, further investigation demonstrated that integrating training data from multiple organisms yields superior predictive power. In a benchmark scenario designed to simulate real-world applications, DeepPEP outperformed the state-of-the-art tool Geptop 2.0, showcasing a robust ability to identify species-specific essential proteins. Finally, a case study on novel genomes confirmed the model's practical effectiveness. Our results suggest that DeepPEP is a powerful strategy for prokaryotic essential protein prediction, and the rigorous evaluation framework established in this study provides a new benchmark for the field.

Large Language Models↗

PubMed2026

An Integrated Machine Learning and Genomic Framework for Precise Detection of Gastric Cancer.

This study presents a novel integrative approach for the analysis of high-dimensional gene expression data, leveraging the complementary strengths of unsupervised clustering and supervised classification. Using K-means clustering, the data set is stratified into three distinct clusters, revealing intrinsic biological patterns and relationships. The resulting cluster assignments are subsequently used as pseudolabels to train machine learning models, including support vector machines, random forest, and a stacking ensemble classifier. To validate and enhance the robustness of clustering, complementary methods, such as hierarchical clustering and density-based spatial clustering of applications with noise (DBSCAN), are used, with results visualized through principal component analysis-driven dimensionality reduction. The high predictive accuracy achieved by the classifiers underlines the separability and reliability of the identified clusters. Furthermore, feature importance analysis highlighted key genetic determinants within each cluster, offering actionable insights into potential biomarkers and critical genomic features. This framework bridges the gap between exploratory unsupervised learning and predictive supervised modeling, providing a scalable and interpretable method for analyzing complex genomic data sets. Its applicability extends to biomarker discovery, patient stratification, and other precision medicine applications, emphasizing its utility in advancing genomic research and clinical practice.

Humans↗

PubMed2025

In silico prediction method for plant Nucleotide-binding leucine-rich repeat- and pathogen effector interactions.

Plant Nucleotide-binding leucine-rich repeat (NLR) proteins play a crucial role in effector recognition and activation of Effector triggered immunity following pathogen infection. Genome sequencing advancements have led to the identification of a myriad of NLRs in numerous agriculturally important plant species. However, deciphering which NLRs recognize specific pathogen effectors remains challenging. Predicting NLR-effector interactions in silico will provide a more targeted approach for experimental validation, critical for elucidating function, and advancing our understanding of NLR-triggered immunity. In this study, NLR-effector protein complex structures were predicted using AlphaFold2-Multimer for all experimentally validated NLR-effector interactions reported in literature. Binding affinities- and energies were predicted using 97 machine learning models from Area-Affinity. We show that AlphaFold2-Multimer predicted structures have acceptable accuracy and can be used to investigate NLR-effector interactions in silico. Binding affinities for 58 NLR-effector complexes ranged between -8.5 and -10.6 log(K), and binding energies between -11.8 and -14.4 kcal/mol-1, depending on the Area-Affinity model used. For 2427 "forced" NLR-effector complexes, these estimates showed larger variability, enabling identification of novel NLR-effector interactions with 99% accuracy using an Ensemble machine learning model. The narrow range of binding energies- and affinities for "true" interactions suggest a specific change in Gibbs free energy, and thus conformational change, is required for NLR activation. This is the first study to provide a method for predicting NLR-effector interactions, applicable to all pathosystems. Finally, the NLR-Effector Interaction Classification (NEIC) resource can streamline research efforts by identifying NLRs important for plant-pathogen resistance, advancing our understanding of plant immunity.

Plant Proteins↗

PubMed2026

EvoSNR-Prom: Predicting promoters at single-nucleotide resolution with label-aware transfer learning of the pretrained EVO model.

The precise identification of promoters is crucial for understanding gene regulation. Deep learning methods have achieved considerable success in promoter prediction, yet most operate at the sequence level with coarse-grained labels. This means they label an entire DNA segment as either a "promoter" or "non-promoter," which results in a lack of the nucleotide-level resolution in prediction. In this study, we propose EvoSNR-Prom, a model designed for promoter prediction at single-nucleotide resolution. EvoSNR-Prom is built on the Evo foundation model and formulates promoter identification as a token-level sequence labeling problem, analogous to named entity recognition in natural language processing. To address the limited contextual information available in single-nucleotide tokenization, we introduce a lexicon-enhanced embedding strategy that incorporates biologically meaningful DNA lexicons, enriching contextual representations and improving the model's ability to capture complex sequence motifs. Furthermore, to enhance predictive performance on small size datasets, we integrate a label-aware transfer learning framework to leverage knowledge from well-annotated source species to a target organism. The results across various prokaryotic datasets show that EvoSNR-Prom achieves excellent performance. This work provides a valuable computational framework for the high-precision analysis of gene regulatory elements, contributing to the advancement of promoter prediction at single-nucleotide resolution.

Promoter Regions, Genetic↗

PubMed2026

shinyDeepGxP: a user-friendly R shiny app for predicting surface protein abundance from scRNA-seq expression using deep learning in blood cells.

MOTIVATION: Understanding accurate immune cell heterogeneity and function in single-cell datasets requires access to protein-level information, which is often unavailable due to experimental limitations. RESULTS: We present shinyDeepGxP, an interactive web application featuring our deep learning model, DeepGxP, for predicting surface protein abundance from single-cell RNA-sequencing (scRNA-seq) data. This platform makes DeepGxP accessible to researchers without programming skills. Users can upload scRNA-seq count matrices and use "Predict Protein" to predict the abundance of 224 biologically relevant surface proteins. shinyDeepGxP provides visualizations to help identify distinct cell populations based on predicted protein profiles. Moreover, users can choose "Explore Model" to reveal key RNA predictors and their associated biological pathways for each protein. Overall, shinyDeepGxP is a user-friendly, freely available web tool that provides protein-level detail for RNA-only single-cell datasets, enabling multimodal discovery without additional experiments. AVAILABILITY AND IMPLEMENTATION: shinyDeepGxP can be launched on https://shiny.crc.pitt.edu/deepgxp/.

Journal Article↗

PubMed2025

Deciphering microbial and metabolic influences in gastrointestinal diseases-unveiling their roles in gastric cancer, colorectal cancer, and inflammatory bowel disease.

INTRODUCTION: Gastrointestinal disorders (GIDs) affect nearly 40% of the global population, with gut microbiome-metabolome interactions playing a crucial role in gastric cancer (GC), colorectal cancer (CRC), and inflammatory bowel disease (IBD). This study aims to investigate how microbial and metabolic alterations contribute to disease development and assess whether biomarkers identified in one disease could potentially be used to predict another, highlighting cross-disease applicability. METHODS: Microbiome and metabolome datasets from Erawijantari et al. (GC: n = 42, Healthy: n = 54), Franzosa et al. (IBD: n = 164, Healthy: n = 56), and Yachida et al. (CRC: n = 150, Healthy: n = 127) were subjected to three machine learning algorithms, eXtreme gradient boosting (XGBoost), Random Forest, and Least Absolute Shrinkage and Selection Operator (LASSO). Feature selection identified microbial and metabolite biomarkers unique to each disease and shared across conditions. A microbial community (MICOM) model simulated gut microbial growth and metabolite fluxes, revealing metabolic differences between healthy and diseased states. Finally, network analysis uncovered metabolite clusters associated with disease traits. RESULTS: Combined machine learning models demonstrated strong predictive performance, with Random Forest achieving the highest Area Under the Curve(AUC) scores for GC(0.94[0.83-1.00]), CRC (0.75[0.62-0.86]), and IBD (0.93[0.86-0.98]). These models were then employed for cross-disease analysis, revealing that models trained on GC data successfully predicted IBD biomarkers, while CRC models predicted GC biomarkers with optimal performance scores. CONCLUSION: These findings emphasize the potential of microbial and metabolic profiling in cross-disease characterization particularly for GIDs, advancing biomarker discovery for improved diagnostics and targeted therapies.

Humans↗

PubMed2025

Predicting natural variation in the yeast phenotypic landscape with machine learning.

Most organismal traits result from the complex interplay of many genetic and environmental factors, making their prediction difficult. Here, we used machine learning (ML) models to explore phenotype predictions for 223 traits measured across 1011 genome-sequenced Saccharomyces cerevisiae strains isolated worldwide. We benchmarked a ML pipeline with multiple linear and non-linear models to predict phenotypes from genotypes and gene expression, and determined gradient boosting machines as the best-performing model. Gene function disruption scores and gene presence/absence emerged as best predictors, suggesting a considerable contribution of the accessory genome in controlling phenotypes. The prediction accuracy broadly varied among phenotypes, with stress resistance being easier to predict compared to growth across nutrients. ML identified relevant genomic features linked to phenotypes, including high-impact variants with established relationships to phenotypes, despite these being rare in the population. Near-perfect accuracies were achieved when other phenomics data mostly in similar conditions were used, suggesting that useful information can be conveyed across phenotypes. Overall, our study underscores the power of ML to interpret the functional outcome of genetic variants.

Genetic Variation↗

PubMed2025

A multi-modal transformer for cell type-agnostic regulatory predictions.

Sequence-based deep learning models have emerged as powerful tools for deciphering the cis-regulatory grammar of the human genome but cannot generalize to unobserved cellular contexts. Here, we present EpiBERT, a multi-modal transformer that learns generalizable representations of genomic sequence and cell type-specific chromatin accessibility through a masked accessibility-based pre-training objective. Following pre-training, EpiBERT can be fine-tuned for gene expression prediction, achieving accuracy comparable to the sequence-only Enformer model, while also being able to generalize to unobserved cell states. The learned representations are interpretable and useful for predicting chromatin accessibility quantitative trait loci (caQTLs), regulatory motifs, and enhancer-gene links. Our work represents a step toward improving the generalization of sequence-based deep neural networks in regulatory genomics.

Humans↗

PubMed1988

Attributional style of depressed chronic low back patients.

This study used the Attributional Style Questionnaire to study the attributional styles of depressed and nondepressed chronic low back pain patients (N = 91) in order to test the Revised Learned Helplessness model's prediction of differences between the two. The results partly supported the hypothesis; an internal, stable, global style for negative events distinguished the depressed group from the nondepressed, but there were no differences in attributional style for positive events. The findings are consistent with recent reviews of the literature that have reported general support for the negative outcome style, but consistent failure to confirm the predictions associated with positive outcome style. In addition, the attributional style was not common to all subjects in the depressed group, which suggested that other factors may be involved in the development of different subtypes of depression. Implications for studying attributional aspects of depression and chronic low back pain are discussed.

Back Pain↗

PubMed2025

Fishing for a reelGene: evaluating gene models with evolution and machine learning.

Assembled genomes and their associated annotations have transformed our study of gene function. However, each new annotated assembly generates new gene models. Inconsistencies between annotations likely arise from biological and technical causes, including pseudogene misclassification, transposon activity, and intron retention from sequencing of unspliced transcripts. To evaluate gene model predictions, we developed reelGene, a pipeline of machine learning models focused on (1) transcription boundaries, (2) mRNA integrity, and (3) protein structure. The first two models leverage sequence characteristics and evolutionary conservation across related taxa to learn the grammar of conserved transcription boundaries and mRNA sequences, while the third uses the conserved evolutionary grammar of protein sequences to predict whether a gene can produce a protein. Evaluating 1.8 million transcript models in Zea mays ssp. mays (maize), reelGene classified 28% as incorrectly annotated or non-functional. We find that reelGene classifies 92.2% of genes in the maize proteome and 99.2% of genes within the maize classical gene list as functional. reelGene also provides a way to further investigate genome biology- for instance, reelGene indicates that 10.3% of dispensable genes in B73 are functional, and within retained duplicate genes, reelGene identifies a 30% bias toward the retention of the M1 subgenome when one copy is functional and the other is non-functional. As an annotation-evaluating tool, reelGene is directly applicable to species of the Andropogoneae tribe, including other important crops like sorghum and miscanthus. As a community resource, reelGene has been integrated onto MaizeGDB both as a browser track and as an individual Shiny App, allowing researchers to evaluate gene model accuracy and further investigate genome biology.

Machine Learning↗

PubMed1975

[Pharmacopsychologic studies on normal subjects for the prediction of the therapeutic efficiency of psychotropic drugs].

The paper deals with some basic problems and possibilities of predicting the therapeutic efficiency of psychotropic drugs from studies in normal humans. Comparing drug studies with normal subjects and patients it seems evident that from a methodological and economical point of view studies with normal subjects have many advantages. However, the practical importance of drug studies with normal subjects is limited unless the therapeutic efficiency of a drug can be predicted. There are some arguments which deny the possibility of prediction, e.g. referring to the lack of comparability of dosages, administration regimens, situational parameters and psychosomatic states between normal subjects and patients. Discussing such arguments it is pointed out that perfect comparability of all these factors is not a necessary prerequisite of prediction. A number of theoretically possible models for predicting therapeutic efficiency is suggested. For some of them there already exists some empirical evidence. The first model takes into account the inter- and intra- individual variability of behavior. It is suggested that a state corresponding to the psychosomatic state of patients be approximated, or simulated, in the normal subjects by suitable selection procedures of subjects or by manipulation of the experimental conditions. The usefulness of such a model for predicting therapeutic efficiency has been shown in many studies with anti-anxiety agents. In another model the drug profiles of normal subjects and patients are compared and the prediction is based upon drug effects in normal subjects which can also be seen in patients and which have a high correlation to the patients' improvement. A further model assumes that improvement is partly the result of learning processes. The prediction of therapeutic efficiency is, therefore, based upon the properties of a drug to facilitate or inhibit learning processes. The final two models proposed, predict limitations of therapeutic efficiency. The one model takes into account side-effects; the other the variability of drug response due to situational and person-parameters.

Humans↗

PubMed2025

Development and Validation an Integrated Deep Learning Model to Assist Eosinophilic Chronic Rhinosinusitis Diagnosis: A Multicenter Study.

BACKGROUND: The assessment of eosinophilic chronic rhinosinusitis (eCRS) lacks accurate non-invasive preoperative prediction methods, relying primarily on invasive histopathological sections. This study aims to use computed tomography (CT) images and clinical parameters to develop an integrated deep learning model for the preoperative identification of eCRS and further explore the biological basis of its predictions. METHODS: A total of 1098 patients with sinus CT images were included from two hospitals and were divided into training, internal, and external test sets. The region of interest of sinus lesions was manually outlined by an experienced radiologist. We utilized three deep learning models (3D-ResNet, 3D-Xception, and HR-Net) to extract features from CT images and calculate deep learning scores. The clinical signature and deep learning score were inputted into a support vector machine for classification. The receiver operating characteristic curve, sensitivity, specificity, and accuracy were used to evaluate the integrated deep learning model. Additionally, proteomic analysis was performed on 34 patients to explore the biological basis of the model's predictions. RESULTS: The area under the curve of the integrated deep learning model to predict eCRS was 0.851 (95% confidence interval [CI]: 0.77-0.93) and 0.821 (95% CI: 0.78-0.86) in the internal and external test sets. Proteomic analysis revealed that in patients predicted to be eCRS, 594 genes were dysregulated, and some of them were associated with pathways and biological processes such as chemokine signaling pathway. CONCLUSIONS: The proposed integrated deep learning model could effectively predict eCRS patients. This study provided a non-invasive way of identifying eCRS to facilitate personalized therapy, which will pave the way toward precision medicine for CRS.

Humans↗

PubMed2025

Integrative multi-omics profiling deciphers tumor microenvironment heterogeneity and immunotherapy vulnerabilities in lung neuroendocrine carcinomas.

INTRODUCTION: Lung neuroendocrine carcinomas (Lu-NECs) are rare, highly aggressive lung tumors with poor prognosis and limited therapeutic options. Understanding the tumor immune microenvironment (TIME) is crucial towards personalized therapeutic strategies. OBJECTIVES: This study aims to systematically characterize the heterogeneity and complexity of the TIME in Lu-NECs by integrating proteomic, transcriptomic, and genomic data. METHODS: We performed comprehensive immune-proteomic profiling of 76 Lu-NECs across diverse histopathological subtypes to elucidate intra-tumoral TIME heterogeneity at the proteomic level. Validation was conducted in multiple independent cohorts, including 112 Lu-NECs using immunohistochemistry, 147 Lu-NECs, and 17 small cell lung carcinoma samples using transcriptomics. We integrated proteomic, transcriptomic, genomic, and clinical data to assess molecular, immunological, and clinical features, as well as therapeutic vulnerabilities across different immune subtypes. RESULTS: We delineated the immuno-proteomic landscape of Lu-NECs and identified two major immuno-proteomic clusters with distinct immunological, molecular, and clinical characteristics. IPC1 was characterized by high immune cell infiltration, while IPC2 exhibited sparse immune cell presence. Genomic analysis revealed distinct mutational patterns, with IPC1 showing a higher incidence of APOBEC-associated mutation signatures and IPC2 being enriched for mutations associated with defective DNA mismatch repair and tobacco-related mutagens. Functional analyses indicated that IPC1 was related to immune and oncogenic signaling activity, whereas IPC2 was associated with cancer stemness and proliferation-related features. Furthermore, IPC1 and IPC2 demonstrated histological subtype-specific clinical benefits from postoperative chemotherapy. Finally, we developed a machine learning model (iPROM) to predict Lu-NECs immune classification and improve risk stratification, which was validated across multiple independent cohorts. CONCLUSIONS: This study advances the understanding of the tumor immune microenvironment in Lu-NECs through multi-omics characterization and highlights potential personalized therapeutic vulnerabilities tailored to the specific immune landscapes of Lu-NECs.

Humans↗

PubMed2026

A voyage of reprogrammable metabolic bioengineering reshapes plant defense: from editing tools to synthetic systems.

Metabolic bioengineering has emerged as a transformative approach for reshaping plant defense by targeting intrinsic biosynthetic pathways to enhance immunity in modern agriculture. Moving beyond proof-of-concept metabolomics to broad-spectrum programmable pathway engineering addresses gaps in plant rational design and optimizes resilience in response to diverse environmental cues. This review aims to comprehensively highlight the transition of innovative approaches to phenolics, alkaloids, flavonoids, terpenoids, and benzoxazinoids, inferring adaptive reprogramming that mediates the growth-defense balance and functions as molecular sentinels in plants. Furthermore, decoding the volatile metabolome reveals a dynamic signaling interface that influences defense responses and stress-induced plant-microbe interactions, with the shikimate, jasmonate, and salicylate pathways functioning as central hubs for microbial deterrence and priming immune memory. Recent developments in multi-scalar genome-editing strategies, including CRISPR-driven combinatorial edits, enzyme orthogonalization, fluxomics, and spatially resolved multi-omics, reconfigure central and specialized metabolic fluxes toward improved defense function and regulation. Additionally, emerging tools, such as WUSCHEL2 and BABY BOOM transcriptional modules, and artificial engineering strategies integrating deep learning model-driven predictions facilitate rapid development of synthetic genetic circuits and support a predictive engineering of plants. Moreover, Mass spectrometry imaging (MSI) in spatial metabolomics enables to obtain structures and locations of unidentified endogenous metabolites within cells and tissues. Overall, this review emphasizes a diverse array of primary and secondary metabolites, spanning molecular concepts to recent advances in plant immune mechanisms. It also illustrates new frontiers in programmable metabolic engineering that accelerate the understanding of plant-microbe-metabolite cross-talks, offering strategies to improve plant resistance and advance sustainable agricultural solutions.

metabolic bioengineering↗

PubMed2025

NRG-P0074 Viral Sample RU1 from Unclassified Mosigvirus Genomic Characterization and Host Range Analysis.

BACKGROUND: Machine learning models for phage-host range prediction and design require comprehensive training data on phage genomes and host ranges to predict phage-host interactions effectively. MATERIALS AND METHODS: This study characterizes phage sample NRG-P0074 viral sample RU1 from unclassified Mosigvirus, originally isolated by the Betty Kutter. The complete genome of NRG-P0074 was sequenced, annotated, and analyzed using various bioinformatic tools. Host range analysis was conducted using the Escherichia coli Reference (ECOR) Library and nine Escherichia coli (E. coli) K12 strains (Keio Knockout Collection) with single nonessential gene deletions. RESULTS: The genome of NRG-P0074 spans 168,357 base pairs with a guanine-cytosine (GC) content of 37.5%. NRG-P0074 exhibited permissiveness in 15.28% of the ECOR isolates and all 9 Keio knockout strains. Comparative genomic analysis revealed that NRG-P0074 is closely related to E. coli phage a20. Its genome is comprised of 270 coding sequences, 153 known genes, 16 terminators, 3 ribosomal-binding sites, 0 tRNAs, and 117 hypothetical proteins. CONCLUSIONS: This research provides valuable data for developing machine learning models to predict phage-host interactions, aiding the development of targeted phage therapies against antibiotic-resistant bacteria.

ECOR Library↗

PubMed1988

From conditioning to category learning: an adaptive network model.

We used adaptive network theory to extend the Rescorla-Wagner (1972) least mean squares (LMS) model of associative learning to phenomena of human learning and judgment. In three experiments subjects learned to categorize hypothetical patients with particular symptom patterns as having certain diseases. When one disease is far more likely than another, the model predicts that subjects will substantially overestimate the diagnosticity of the more valid symptom for the rare disease. The results of Experiments 1 and 2 provide clear support for this prediction in contradistinction to predictions from probability matching, exemplar retrieval, or simple prototype learning models. Experiment 3 contrasted the adaptive network model with one predicting pattern-probability matching when patients always had four symptoms (chosen from four opponent pairs) rather than the presence or absence of each of four symptoms, as in Experiment 1. The results again support the Rescorla-Wagner LMS learning rule as embedded within an adaptive network model.

Association Learning↗

PubMed2025

Integration of Gene Expression and Digital Histology to Predict Treatment-Specific Responses in Breast Cancer.

Deep learning models applied to digital histology can predict gene expression signatures (GES) and offer a low-cost, rapidly available alternative to molecular testing at the time of diagnosis. We optimized transformer-based models to infer GES results and applied this approach to pre-treatment H&E-stained biopsies from 1,940 breast cancer patients treated with neoadjuvant chemotherapy in clinical trial and real-world cohorts. The most predictive histology-derived GES for pathologic complete response (pCR) in the I-SPY2 trial was validated in four external cohorts: CALGB 40601, CALGB 40603, a trial of durvalumab plus CT, and standard-of-care CT-treated patients from the University of Chicago. Among HER2-negative patients, a transformer-based model trained using a signature composed of estrogen-regulated genes, proliferation, apoptosis, and interferon response genes predicted pCR with an AUC of 0.794, outperforming models based on clinical features alone (AUC 0.704, p = 0.001), pathologist TIL assessment, and a model trained directly to predict response from I-SPY2 cases. Tertiles of this signature stratify patients into clinically relevant groups with increasing likelihood of complete response, with pCR rates ≥50% in the top tertile regardless of treatment or hormone receptor status. Additional transformer-based signature models predicted response to specific therapies (but not chemotherapy alone), including a HER2 signaling signature in IO-treated patients, and a claudin-low signature in bevacizumab treated patients. In HER2- cohorts with available gene expression data and histology, models trained on expression data performed similarly to digital histology predictions, but the combination of gene expression and histology outperformed histology alone. These findings suggest that histology-based GES provides additive information to RNA sequencing data and can inform precision treatment selection across breast cancer subtypes.

Journal Article↗